T-network: the better feedback solution?

[chris ma]

I mean, that many problem here rise, because not exist common listening experience, mostly it is only one subjective statement against another subjective statement. All we can to get best sound, but experiences and knowledges aren't the same. I have experience from our czech chat, that people often don't belive claims, but when they comes together for listening, in most cases is 100 % agreement. It should not to be easy, but whatabout to do some " listening meeting ", where we can try it together ? But there must be some different products, not only variation of only one idea - any fan club of one solution. 😎

I agree 100%; people keep upgrade their equipments without updating the listening experience always bother me a bit. I myself included. But there are not many good systems to base on except a few local hi-fi shops that are normally not that friendly when they suspect that you have no intention to buy. Group meet is a good idea.

The Butcher😀


PS : you never know the best until you get to know the second best!

 

[TwoSpoons]

It is not very exact to explain the sound by words, as it's always very difficult and you are picking words that are valid for yourself, they are obvious for you but not so for others.
I read very funny words on the audio mags.😀

Wild use of adjectives to try to convey what is essentially a highly subjective experience, is a constant source of annoyance to me.

What the hell does 'smoother' or 'balanced' or 'detailed' actually mean in real terms? Frankly its all BS until quantified by some real measurements i.e. gain and phase. What I would like to see is a description like 'smoother' backed up with before and after gain/phase plots, so we can get some idea of what is happening, and replicate the results, or better still get some idea of how to reach a particular design goal.

Err ... this reads a bit like a dig at carlosfm 😱 . Its not - I have great respect for your work carlos, and I'll definitely be adding the T-net to my design.

 

[planet10]

What I would like to see is a description like 'smoother' backed up with before and after gain/phase plots, so we can get some idea of what is happening

Those kinds of things won't tell you much, you need to examine some sort of measure that tells you what is happening 40-50 dB down from the main signal which includes full spectra of distortion harmonics, and more.... no one has invented these tests yet -- at least that i know of, so we have to rely on our highly evolved (and hopefully trained) ear-brain system.

dave

 

[carlosfm]

Err ... this reads a bit like a dig at carlosfm 😱 . Its not - I have great respect for your work carlos, and I'll definitely be adding the T-net to my design.

I don't read your post as a "dig" at me, as I was saying almost the same thing.
Although in my oppinion you can't quantify everything by numbers, they may give you a general idea, but it's not a real picture of what you will listen.

BTW, T-network applied to the GC was Franz's idea, not mine.

 

[Pedja]

I think Joe was in that for some time before Franz.

 

[carlosfm]

I think Joe was in that for some time before Franz.

No, Pedja.
The discussion started on Joe's thread, but it was brought in by Franz.
Then Franz opened this thread.
Joe tried it and reported good results.

EDIT: of course, some credit must go to Joe by being one of the first (if not the first, not sure) to try it, and suggesting some component values.

 

[Pedja]

Yes, Carlos, I’ve checked it back, you are right. I probably had that impression because Franz looked like he came to the board to get the help sorting the circuit, and Joe was very fast providing necessary info.

Pedja

 

[Franz G]

The discussion started on Joe's thread, but it was brought in by Franz.
Then Franz opened this thread.
Joe tried it and reported good results.

EDIT: of course, some credit must go to Joe by being one of the first (if not the first, not sure) to try it, and suggesting some component values.

Exactly, Carlos!

I realized with my first tube buffered amp, that Joe is using a problematic high value for feedback and asked him why. He never answered.

Later, I have seen this t-net topology in opamp theory and immediately began to play with it. The result was so amazing, I quickly posted the t-net in the VBIGC thread.

I tried it before Joe, with an regulated OPA627 buffered Amp, but obviously at least one gaincloner used this technique one or two years before, but did not call it t-net :-(.

And: from the beginning, I offered an excel sheet to calculate the values for the t-net and the gain.

It is a good example, not having some rights on this development: the t-net is not my invention, it is very old.

So, I was me, sucessfully "selling" the t-net idea to the forum, assisted by Joe and some other guys, testing the t-net quickly.

The nice fact: it sounds much better!

Franz

 

[Franz G]

and Joe was very fast providing necessary info

As you can see now, Joe was working hardly at the JLTI MKII.

And because the t-net is so much better, he just was in time to pick it up quickly!

Imagine what would happen to a MKII without t-net, while we are all t-networking! (maybe a MKIIa or better a MKIIt?)

Franz

 

[theChris]

"EDIT: of course, some credit must go to Joe by being one of the first (if not the first, not sure) to try it, and suggesting some component values."
-not sure when joe did it. i did it in january, 2004, and had it deigned a few months before, also with good results. it was posted and got a little interest. I called it High Input Impedance Inverting GainClone - hi3gc, a name that obviously never caught!

i also tried adapting the circuit to a Non-Inverting amp. the result was not suprising. if the user can deal with a higher DC offset, the feedback capacitor can be made smaller without affecting the cutoff frequency.

i'll try to find out what components i used. i vaugely remember it lined up right in terms of biasing and gain, but had odd values that i had laying around.

 

[Franz G]

but obviously at least one gaincloner used this technique one or two years before, but did not call it t-net :-(.

theChris was this mentioned gaincloner!

Franz

 

[Sheldon]

Re: Re: Puzzling problem, experts needed

Well, another bit of info, and a solution. But I still don't understand it.

The new info is that the problem does not occur if the buffer is not connected when the main ps is turned on (remember the buffer is always on, ala Joe R's tube buffer). If I turn on the ps and then connect the buffer, it acts normally. The offset is even a little lower. If I mute the amp during turn on, no problem. I had thought it unnecessary, at least for the turn on thump. Not so, at least with this chip. Some sort of weird latch up (one one channel only and only with one of the buffers) if the buffer is on while the amp turns on.

Back to see what I can do with the NE555 and a relay.

Sheldon

Mystery solved. I had thought I had a bad amp chip. Not so. I turns out that the coupling cap in the buffer had a poor internal connection and was probably shorting and passing some DC to the input which the amp amplified (as it's supposed to). The cap finally failed as an open circuit and I got no sound at all. Replaced cap, things work.

Sheldon

 

[Franz G]

As I always solder the t-nets under the board, I began to mold them into epoxy. And this looks like this:

An externally hosted image should be here but it was not working when we last tested it.

Franz

 

[carlosfm]

Visitors from Mars?😀

 

[yldouright]

The ITGC?

This thread needs resussitation. I want to build a T-Network Inverted Gainclone (hereafter we could call it the ITGC if Franz and theChris don't object) as an integrated amp. I want to make it for a battery supply first (+/-12V) and eventually try using the T-Network on a gainclone circuit with differential noise reduction. I was wondering if using a linear pot with a law faking resistor would mitigate problems for the sources. Would someone simulate the ITGC with a 100K linear pot parallel with something like a 33K resistor to see if this combination adequately handles source inputs with 1K output impedences? I'm hoping to avoid using a buffer without giving up any sonics.

 

[theChris]

how about T-GC? or GTC? GTC is kinda catchy. i called mine HI3GC (high input impedance inverting gainclone) because i wanted to use a small 1uF mylar cap instead of larger electrolytics. i could live with ITGC as well.

the closest T clone for non-inverting i've thought up looks like an H, but an H-Clone makes you think its class H...

note that for an IGC of any type the input resistors are in parallel with the input as seen by the source, thus you may be able to get away with just the pot and the normal input resistor. just make the input resistor 33k.

 

[Greg Erskine]

hi theChris,

I just tried to find your original post but to no avail. I was wondering if you could let me know the values of the feedback resistor network and the input resistor.

I am curently using 22k input, 10k-100R-10k feedback resistors with NO input cap, buffer or filter. I have hiss, especially from the tweeters. I have tried swapping speakers, power supply and finally put my original IGC back, so I have narrowed it down to my bad implementation of the TIGC.

From my readings, I worked out resistors should be around 22k, 10k-10k-330k which is so different from other recommendations that it must be WRONG!

Thanks

 

[yldouright]

grege
I suppose we could use "T" as the prefix the nomenclature for this topology but I think it might be less confusing to put it directly before the GC acronym as theChris suggests so that TGC and NITGC are t-network gainclones. If we use T as the prefix we may end up with a VBTNIGC!?

As a suggestion to the hiss problem you can try several things:

(a) reduce the 330KΩ resistors to 220KΩ
(b) increase the 100Ω resistors to 220Ω
(c) use another chip

I would try each of the above suggestions individually so there are at least eight variations for a solution. If you built the ITGC P2P then this is going to be a pain but I don't think you can do otherwise.

 

[Greg Erskine]

I suppose we could use "T" as the prefix the nomenclature for this topology but I think it might be less confusing to put it directly before the GC acronym as theChris suggests so that TGC and NITGC are t-network gainclones. If we use T as the prefix we may end up with a VBTNIGC!?

thanks yldouright

MY logic dictates that the "T" needs to be a prefix because "I" and "NI" are just about always used, ie IGC or NIGC, so why change because you added a T-network. Its just like adding the "VB", so VBTNIGC looks fine to me. 😀 How about this one, RVBSLCTIGC = regulated valve buffered snubbered large cap t-network inverted gainclone.

I have changed my TIGC or ITGC roughly to Carlosfm's suggestion (15k, 18k-1-18k) but seem to have destroyed it in the process, so I don't know if would have worked. I have rail voltage on the output and can't see any shorts. So it looks like testing T-networks is on hold till I feel like building another one.

 

[yldouright]

grege
Using "TNIGC" may be interpreted as a "T-Network" Inverted Gainclone but you're right, either designation is functional. My vote goes for the "I" before "T" but I'll comply with whatever we agree on. Sorry to hear about your gainclone ca"T"astrophe but are you sure you understood carlosfm correctly? Those values don't even look functional to me. Have you tried to use the FranzG spreadsheet calculator? Try plugging the values you used into it and see what you get.
🙂